15674
Biochemistry 2000, 39, 15674-15685
Articles Analysis of Antibody A6 Binding to the Extracellular Interferon γ Receptor R-Chain by Alanine-Scanning Mutagenesis and Random Mutagenesis with Phage Display† Stefan Lang,‡ Jian Xu,‡ Fiona Stuart,‡ Richard M. Thomas,§ Jan W. Vrijbloed,‡ and John A. Robinson*,‡ Institute of Organic Chemistry, UniVersity of Zu¨ rich, Winterthurerstrasse 190, 8057 Zu¨ rich, and Department of Polymer Chemistry, ETH-Zurich, Switzerland ReceiVed April 13, 2000; ReVised Manuscript ReceiVed August 3, 2000
ABSTRACT: The monoclonal antibody A6 binds a conformational epitope comprising mainly the CC′ surface loop on the N-terminal fibronectin type-III domain of the extracellular interferon γ receptor (IFNγR). The crystal structure of an A6 Fab-IFNγR complex revealed an interface rich in the aromatic side chains of Trp, Tyr, and His residues. These aromatic side chains appear to interact with both polar and hydrophobic groups at the interface, a property which, in general, may be advantageous for ligand binding. To analyze these interactions in more detail, the affinities of 19 A6 alanine-scanning mutants for the IFNγR have been measured, using engineered A6 single chain variable region fragments, and a surface plasmon resonance biosensor. Energetically important side chains (∆Gmutant - ∆Gwt > 2.4 kcal/mol), that form distinct hot spots in the binding interface, have been identified on both proteins. These include VLW92 in A6, whose benzenoid ring appears well situated for a π-cation (or π-amine) interaction with the side chain of receptor residue K47 and simultaneously for T-stacking onto the indole ring of W82 in the receptor. At another site, energetically important residues VHW52 and VHW53, as well as VHD54 and VHD56, surround the aliphatic side chain of the hot receptor residue K52. Taken together, the results show that side chains distributed across the interface, including many aromatic ones, make key energetic contributions to binding. In addition, the receptor CC′ loop has been subjected to random mutagenesis, and receptor mutants with high affinity for A6 have been selected by phage display. Residues previously identified as important for receptor binding to A6 were conserved in the clones isolated. Some mutants, however, showed a much improved affinity for A6, due to changes at Glu55, a residue that appeared to be energetically unimportant for binding the antibody by alanine-scanning mutagenesis. An E55P receptor mutant bound A6 with a 600-fold increase in affinity (KD ≈ 20 pM), which is one of the largest improvements in affinity from a single point mutation reported so far at any protein-protein interface.
The extracellular portion of the interferon γ receptor R-chain (1) (IFNγR)1 comprises two FBN-III domains that together are responsible for binding IFNγ at the cell surface. The crystal structure of a complex formed between the neutralizing murine mAb A6 and the N-terminal FBN-III † The authors thanks the Swiss National Science Foundation for financial support. * To whom correspondence should be addressed. Phone: (++)-411-635-4242. Fax: (++)-41-1-635-6812. E-mail:
[email protected]. ‡ University of Zu ¨ rich. S.L., J.X., F.S., and J.V. all contributed equally to this work. § Department of Polymer Chemistry. 1 Abbreviations: amp, ampicillin; CDR, complementarity determining region; FBN-III, fibronectin type-III domain; IFNγ, human interferon γ; IFNγR, extracellular portion (residues 1-229) of the human interferon γ receptor R-chain; mAb, monoclonal antibody; nt, nucleotides(s); scFv, single chain variable region fragment [VH + VL joined by a (G4S)3 linker]; tet, tetracycline; Trx, E. coli thioredoxin; VH, heavy chain variable region (residues are labeled, e.g., VHW96); VL, light chain variable region (residues are labeled, e.g., VLR108); wt, wild-type.
domain showed that the surface of the receptor contacted by A6 overlaps significantly with that bound by IFNγ in the ligand-receptor complex (2, 3). The A6 epitope encompasses largely the CC′ hairpin loop on the receptor, although contact is also made with the neighboring F-strand (Figure 1). The antigen-antibody interface buries ca. 1660 Å2 of protein surface, which includes 22 antibody residues from 5 CDRs and 11 residues from the receptor. Antibody-antigen complexes remain interesting targets for mechanistic studies of protein-protein recognition, not least because high-affinity protein-protein interactions underlie many important processes in biology. Such studies are also of interest in the context of designing proteins, and smaller molecules, with novel recognition sites. One feature of interest in this case is the high concentration of aromatic groups (six Tyr, six Trp, and one His) located at the A6IFNγR interface (four in VL, six in VH, and three in the receptor). This reflects a broader tendency for aromatic
10.1021/bi000838z CCC: $19.00 © 2000 American Chemical Society Published on Web 12/01/2000
Mutagenesis of A6 Antibody
FIGURE 1: Ribbon diagram of the IFNγR1-108-A6 Fab complex (2). The Fv (VH + VL) region of A6 is shown complexed to the N-terminal FBN-III domain of the receptor (PDB file 1JRH).
groups to be over-represented statistically in antigen binding sites on antibodies (4, 5) and also, more generally, at protein-protein interfaces (6-8). The crystal structure of the A6-IFNγR complex reveals several aromatic rings which might participate in multiple types of interaction simultaneously, a property that may be relevant to their frequent occurrence in antigen binding sites. The energetic importance of such interactions is, therefore, of considerable interest. We describe here new data on the contribution to binding affinity of antibody side chains buried at the A6-IFNγR interface, which complements earlier studies involving mutagenesis of the receptor (9). An experimental system for the study of an A6 scFv fragment is described, together with affinity measurements of scFv alanine-scanning mutants (10) using a surface plasmon resonance biosensor (BIACORE, Biacore AB). In addition, the receptor CC′ loop has been subjected to random mutagenesis and receptor mutants have been selected by phage display that have almost 3 orders of magnitude higher affinity for A6 than the wt. In particular, an E55P mutant has an approximately 600-fold higher affinity for the antibody, although the residue Glu55 appeared by alanine-scanning mutagenesis to be energetically unimportant for binding to the antibody. EXPERIMENTAL SECTION General. DNA and protein manipulations were as described in ref 11. Production of IFNγR with an Engineered C-Terminal Cys. The phagemid pOCI656 described earlier (2) was used as template DNA in a PCR with the primers: left 1, 5′GGATTGTTATTACTCGCGGC-3′; right 1: 5′-TTCTGCGGCCGCCTAACAAGAACCTTTTATACTGC-3′. The left 1 primer binds to a sequence upstream of the SfiI and NcoI sites, present also in the parent phagemid pHEN1 (12). The right 1 primer (NotI site underlined; Cys and stop codons in italics) encodes the C-terminus of the receptor on the complementary strand, i.e., ...Ser224-Ser225-Ile226-Lys227Gly228-Ser229-Cys-stop. When the PelB signal peptide is cleaved, the receptor from Glu1-Met2-Gly3... to Ser229 with an additional Cys residue after Ser229 is produced and secreted into the Escherichia coli periplasm. The PCR product was digested with NotI and NcoI and cloned into NcoI/NotI digested pHEN1 to give pOCI1110. The nt sequence of the insert was confirmed by DNA sequencing.
Biochemistry, Vol. 39, No. 51, 2000 15675 Plasmid pOCI1110 was introduced into E. coli HB2151. The recombinant receptor was produced using a method described earlier (2), except that after the periplasmic extraction, β-mercaptoethanol (5 mM) was added to all buffers, and purification included immunoaffinity chromatography with antibody A6 coupled to CNBr-activated Sepharose (Pharmacia). After purification, the receptor was homogeneous by SDS-PAGE, gave an electrospray mass spectrum consistent with the calculated mass (25 970 Da), and the correct N-terminal sequence by Edman degradation. The protein was stable for at least 2 months at 4 °C in BIACORE buffer [sodium phosphate (50 mM), pH 7.2, NaCl (150 mM) and 0.005% (v/v) surfactant p20] containing β-mercaptoethanol (5 mM) and EDTA (5 mM). Upon prolonged storage, proteolysis led to the partial loss of the five N-terminal residues. This did not appear to affect recognition by A6. Production and Characterization of A6 wt scFV. The primers left 1 and right 2 were used in a PCR with phagemid pOCI455 (13) as template: right 2, 5′-CCGGAATTCGCGGCCGCTTACCGTTTCAGCTCCAGCTTGGTC-3′. The primer right 2 encodes, on the complementary strand, the C-terminus of VL up to VLR108 (14) followed immediately by a stop codon (underlined) and further downstream an EcoRI site (underlined). The PCR product was digested with SfiI and EcoRI and cloned between the same sites in the vector pHB110 (15) to produce pOCI1103. This encodes a PelB leader sequence fused to the N-terminus of VH, followed by the (G4S)3 linker, and VL up to and including VLR108. The leader peptide is cleaved during transport into the periplasm to give A6 scFv with residues Q1V2K3... at the N-terminus. The A6 scFv was produced by minor modifications to methods described earlier (13). The plasmid pOCI1103 was introduced into E. coli HB2151. The IPTG concentration for induction was 1 mM, and after induction, growth was at 26 °C for ca. 16 h. A periplasmic extract was made using a method described earlier for production of the IFNγR (2). The A6 scFv was purified by immunoaffinity chromatography with Trx-IFNγR1-108 (16) coupled to CNBr-activated Sepharose, as described earlier (13). The protein from this column was applied to a MonoS ion-exchange column (Pharmacia) equilibrated in sodium phosphate buffer (50 mM, pH 6.8). The nativelike scFv eluted in the flow-through, whereas impurities could be eluted from the column in a salt gradient (0 to 0.5 M NaCl). The resulting protein showed a single band by SDS-PAGE, gave an electrospray mass spectrum consistent with the calculated mass, and the correct N-terminal sequence by Edman degradation. The concentration of A6 scFv in PBS buffer [sodium phosphate 50 mM, NaCl (150 mM), pH 7.4] was determined by absorption at 280 nm (A280), using A280 ) 2.9 mg mL-1 after calibration by quantitative amino acid analysis. Gel filtration chromatography of A6 scFv was in PBS buffer on a TSK-GEL G3000SW column (7.5 × 600 mm) (Tosohass, Stuttgart) at a flow rate of 0.7 mL/min. CD spectra were recorded in PBS buffer using a Jasco J720 instrument at 25 °C. Analytical ultracentrifugation was performed using a Beckman XL-A instrument at 5 °C and 25 °C, at both 0.5 and 5 µM protein concentration in PBS buffer, at a rotor speed of 22 000 rpm. Data were acquired by averaging 20 radial scans at a spacing of 0.001 cm, and analyzed assuming
15676 Biochemistry, Vol. 39, No. 51, 2000
Lang et al.
a single homogeneous species according to (eq 1)
cr ) co exp[Mr(1 - νF)ω (r - ro )]/2RT 2
2
2
(1)
where cr is the concentration of the protein at a given radial position, co is the concentration at a reference position, Mr is the apparent molecular weight, ν is the partial specific volume, F is the solvent density, ω is the angular velocity, r and ro are radial positions (in cm) at an arbitrary position and at the meniscus, respectively, R is the gas constant, and T the absolute temperature. Production of Mutant scFV. Mutations were introduced into the cDNA encoding the scFv by PCR with a mutagenic primer, using the method described earlier for mutagenesis of the receptor (9). The mutant scFvs were produced and purified by the same method used for the wt protein (see above), but also included gel filtration chromatography (see above). Each mutant showed one band of ca. 26 kDa on SDS-PAGE. CD spectra were recorded as for the wt (see above). BIACORE Binding Assays. The IFNγR1-229 containing a C-terminal c-myc tag, produced and purified as described earlier (9), was immobilized on a CM5 sensor surface (Biacore AB) by random amine coupling, using a method described earlier (2). The IFNγR1-229 with an engineered Cys residue near the C-terminus (see above) was immobilized on CM5 sensor surfaces by thiol exchange. The surface was activated by injection of N-hydroxysuccinimide (50 mM) and N-ethylN′-(dimethylaminopropyl)carbodiimide (200 mM) in water for 3 min, followed by 2-(2-pyridinyldithio)ethylamine (80 mM) in borate buffer (0.1 M, pH 8.5) for 1 min, and then Cys-containing IFNγR1-229 (20 µg/mL) in sodium acetate buffer (10 mM, pH 4.0) for 2 min. Excess disulfides were blocked by injecting cysteine (50 mM) in formate buffer (0.1 M, pH 4.3) containing NaCl (1 M). Binding curves were measured in BIACORE buffer at 25 °C. Association constants (KA) were determined under equilibrium conditions from Scatchard plots according to the following equation (17, 18):
Req/C ) KA(Rmax - Req)
(2)
where Req is the steady-state response, C is the protein concentration in the flow buffer, and Rmax is the maximum surface binding capacity. KA can be derived from the gradient of a plot of Req/C versus Req at different analyte concentrations (C). The standard deviation of the KA determination upon repeated measurement with the same protein sample was typically (5%. For kinetic analyses, an equation for a simple 1:1 interaction (eq 3) was fitted globally to binding curves collected at various analyte concentrations using the BIAevaluation software (v.3.0) (Biacore AB) (19):
dR/dt ) konC(Rmax - R) - koffR
(3)
where R is the response in RU and kon and koff are the association and dissociation rate constants. Construction of pOCI1009. A PCR reaction was performed using the oligonucleotides primer 1 and primer 2 (see Table 1 in Supporting Information) and pOCI656 (2) as template DNA. A DNA fragment of the expected size (ca. 300 bp) was isolated by gel electrophoresis, digested with NotI and
Table 1: Alanine-Scanning Mutants Studied in This Work, the Measured Dissociation Constants (KD), the Ratio of Kds for Mutant and wt (∆∆G ) RT ln(Kmut/Kwt)) mutant
KD (nM)
KD(mut)/ KD(wt)
∆∆G (kcal/mol)
wild-type VLE27A VLD28A VLY30A VLY91A VLW92A VLS93A VLT94A VLW96A VHY32A VHW52A VHW53A VHD54A VHD55A VHD56A VHY58A VHR95A VHF98A VHY99A VHH100bA
12 ( 0.3 30 ( 0.9 25 ( 0.4 78 ( 7 32 ( 1.5 1400 ( 80 4 ( 0.2 23 ( 0.5 200 ( 100 135 ( 6 1122 ( 35 716 ( 50 290 ( 30 200 ( 50 275 ( 5 100 ( 3 30 ( 2 12 ( 0.3 72 ( 5 211 ( 5
2.5 2.1 6.5 2.7 117 0.33 1.9 17 11.3 94 60 24 17 23 8.3 2.5 1.0 6.0 18
0.54 0.44 1.1 0.58 2.8 -0.65 0.38 1.7 1.4 2.7 2.4 1.9 1.7 1.8 1.2 0.54 0 1.1 1.7
ClaI and subsequently ligated in NotI and ClaI digested pOCI656, to afford pOCI1009 (Figure 2). The nt sequence was confirmed by DNA sequencing. Preparation of Control Phage. A tetracycline resistance gene was excised from pBR322 on a 1.4 kb EcoRI-AVaI fragment and the 5′ ends were blunted using the Klenow fragment of PolI. The fragment was ligated in DraI digested pHEN1 (12) to afford pOCI1005. The ampicillin (amp) resistance gene of pHEN1 is removed from the plasmid by the DraI digestion. ELISA Phage Panning. The method for phage panning was modified from (20). ELISA plates were coated overnight at 4 °C with mAb A6 (200 µL of 10 µg/mL). After washing three times with phosphate buffered saline [PBS; sodium phosphate, 50 mM, pH 7.2, and NaCl (150 mM)], the wells were incubated with PBS containing 1% BSA (200 µL) and incubated 2 h at 37 °C. The ELISA plate was washed again three times with PBS and finally the phagemid particles (ca. 109) were added in PBS containing 1% BSA and the plate was incubated overnight at 4 °C. The wells were washed 25 times with PBS containing 0.1% Tween 20, and 15 times with PBS, incubated 30 min at 20 °C, and finally washed 10 times with PBS. Bound phagemid particles were eluted with glycine buffer (200 µL, 50 mM, pH 2.5). The eluted material was neutralized with phosphate buffer (2 M, pH 7.5) and used for transfection of logarithmic growing E. coli XL1-blue cells. Library Construction. A PCR was performed with the oligonucleotides primer 3 and primer 4 (see Table 1 in Supporting Information) and pOCI1009 as template. The product of the PCR was gel purified and used together with primer 5 in a second PCR again using pOCI1009 as template DNA. The resulting PCR product (ca. 800 bp) was gel purified, digested with NcoI, NotI, and cloned in NcoI/NotI digested pOCI1009, yielding plasmid pOCI1010 in which the cDNA encoding the central portion of the CC′ receptor loop has been deleted (Figure 2). To construct the libraries, cassettes were prepared by annealing oligo 1 with oligo 2, oligo 3 with oligo 4, and
Mutagenesis of A6 Antibody
FIGURE 2: Phagemids pOCI656 and pOCI1010. Below is shown the nt sequence of the CC′ loop region in the wt state (in pOCI1009), and after modification as in pOCI1010. The dotted lines indicate corresponding positions in the nt sequence in each clone, and show how the segment encoding the central part of the loop was deleted in pOCI1010 (see Experimental Section and Results). Oligonucleotide cassettes encoding the mutant libraries were ligated into Asp718/MluI digested pOCI1010.
oligo 5 with oligo 6 (see Table 1 in Supporting Information). A total of 31.9 pmol of each of the six oligonucleotides was annealed in the appropriate combinations by 5 min heating at 95 °C and slowly cooling to 30 °C (30 min). The complementary strands were synthesized by extension of the 3′ ends by adding 1 mM dNTP and 20 units of klenow fragment of PolI and continued incubation at 30 °C for 1 h. The resulting double-stranded DNA was digested with Asp718 and MluI and then ligated in Asp718 and MluI digested pOCI1010. Introduction of the ligated DNA into E. coli XL1-blue cells was by electroporation using a genepulser (Bio-Rad). Colonies were pooled in fractions and frozen in 20% glycerol. Preparation of Phage. Frozen stocks were used to inoculate 2YT (25 mL) containing 1% glucose and ampicillin (100 µg/mL). The culture was grown for 4 h at 37 °C and the cells were pelleted by centrifugation (4000 rpm) and washed once with 2YT. The cells were resuspended in 2YT
Biochemistry, Vol. 39, No. 51, 2000 15677 and M13KO7 phage was added at a multiplicity of infection of 40. After 20 min incubation at 37 °C without shaking, incubation was continued for 1 h at 37 °C with shaking. Then kanamycin (20 µg/mL) was added to the culture, which was 20-fold diluted in fresh 2YT containing ampicillin (100 µg/ mL) and kanamycin (20 µg/mL) and grown overnight at 37 °C. Glycine-eluted phage obtained after a round of panning was directly used to infect logarithmic growing E. coli XL1blue. The cells were pregrown on 2YT containing 1% glucose to an OD600 of 1.0. After 20 min incubation at 37 °C without shaking, incubation was continued at 37 °C with shaking for 1 h. At this point ampicillin (100 µg/mL) was added to the culture and incubation was continued for 4 h at 37 °C. Infected cells were harvested by centrifugation at 4000 rpm and the cell pellet was resuspended in 2YT containing ampicillin (100 µg/mL). Infection by M13KO7 phage was as described above. Phagemid particles from liquid cultures were obtained as described by (21). Production of Mutant Receptor Proteins. The E55P mutant IFNγR1-229 cDNA was prepared by site-specific mutagenesis and PCR using methods described elsewhere (9). Production and purification of wt IFNγR and mutant proteins, each fused to a C-terminal c-myc tag, was as described previously (2). For competition BIACORE, proteins were produced that lacked the c-myc tag, and these were purified by immunoaffinity chromatography with A6. The homogeneity of purified proteins were confirmed (>95%) by SDS-PAGE, and the masses by electrospray mass spectrometry. Competition ELISA. KDs of the wt IFNγR and mutant proteins binding to A6 were determined by competition ELISA as described earlier (9). Competition BIACORE Measurements. The wt IFNγR1-229 was immobilized on CM5 sensor surfaces by random amine coupling, as described earlier (2). The level of immobilized receptor corresponded to g10 000 RU. The wt A6 scFv together with various concentrations of wt IFNγR1-229 were preincubated for at least 1 h at 4 °C, before injection over the IFNγR1-229-sensor surface (flow rate 25 µL/min, injection volume 150 µL). At least 10 binding curves with different concentrations of IFNγR1-229 were measured. Each measurement was performed at least in duplicate. The slopes of the association phases were determined using BIAevaluation, software v.3.0, and were plotted against the total antigen concentration. The KD was determined by fitting to eq 4 (22) using KALEIDAGRAPH (Synergy Software, PA):
(
robs ) rmax 1 -
x(
(
KD + [Agtot] + [Abtot] 1 ‚ 2 [Abtot]
)
KD + [Agtot] + [Abtot] 2 - [Agtot][Abtot] 2
))
(4)
where robs is the slope at a given [Agtot], rmax is the maximal slope in the absence of inhibition by co-incubated antigen, [Abtot] is the total scFv concentration, [Agtot] is the total antigen concentration, and KD is the dissociation constant. RESULTS Expression of Extracellular IFNγR. The wt IFNγR, residues 1-229 (IFNγR1-229), fused at the C-terminus via a short -Ala3- linker to a c-myc peptide tag (23), was produced
15678 Biochemistry, Vol. 39, No. 51, 2000 in E. coli using a secretion expression system described earlier (2). IFNγR1-229 with a cysteine residue engineered at the C-terminus (but without the c-myc tag) was produced in the same way. Both proteins were purified by immunoaffinity chromatography, using either immobilized A6 mAb or the 9E10 mAb against the c-myc tag (23). Expression and Properties of A6 scFV. The A6 scFv fragment comprising the VH domain connected through a linker [(G4S)3] to the VL chain with a c-myc tag attached to the C-terminus (yield ≈ 0.3 mg/L), was produced and purified as described earlier (13). However, the yields of some A6 mutants made in this way were much lower, so an improved expression system was implemented. The wt A6 scFv coding region was cloned into vector pHB110, which also contains the E. coli skp gene (15). After purification, the scFv was obtained in a yield of ∼1.2 mg/L. Eleven of the 19 mutants were obtained in the same way in about the same yield as wt, the VHH100bA, VLW92A, VLY91A, VHD54A, and VLW96A mutants were obtained in 5-6-fold lower yield, and the VHW53A, VHW52A and VHD55A mutants in ∼10-fold lower yields. When micromolar solutions of wt A6 scFv were diluted to 100 nM and injected over a biosensor surface displaying immobilized IFNγR, the shape of the binding curve was seen to change with elapsed time (over ∼60 min) after dilution from the stock solution (data not shown). Freshly diluted sample gave a pronounced biphasic binding curve with a significant contribution from a component with a very slow dissociation rate. The binding curves at later times resembled those typical for a 1:1 interaction with a fast dissociating species. Such behavior could be explained by a dynamic equilibrium between monomeric and multimeric forms of A6 scFv. Upon sedimentation equilibrium analytical ultracentrifugation, the fit of a model which assumed a single molecular species to the sedimentation data was influenced by the temperature and the protein concentration. At a higher protein concentration (5 µM) and 5 °C, the apparent molecular weight was higher than expected and the fitting was poor. A 10-fold dilution to 0.5 µM and measurement at 25 °C, however, gave a molecular weight determination (27.9 kDa) close to the calculated mass with good fitting of the model (Figure 3). Gel filtration chromatography of A6 scFv at high protein concentrations also indicated the presence of dimer, which disappeared from chromatograms upon dilution of the scFv to
